An international research team has developed an in silico method that uses a drug's interaction with proteins to predict the possibility of an adverse event as well as to identify potential new uses for existing treatments.
Investigators from China's Shanghai Jiao Tong University, the US Food and Drug Administration, and elsewhere used the method to develop the Drug Repositioning Potential and Adverse Drug Reaction via the Chemical-Protein Interactome, or DRAR-CPI, web server, which uses drug-drug associations, as well as interactions with proteins to suggest new targets for old drugs and identify unexpected adverse reactions.
In an article describing the method that was published earlier this year in Nucleic Acids Research, the team explained that because "both new indications and [adverse drug reactions] are caused by unexpected chemical-protein interactions on off-targets, it is reasonable to predict these interactions by mining the chemical-protein interactome."
The server, which is available here, currently contains 385 structural models of targetable human proteins and 254 small molecules with known indications and adverse events, including antipsychotics, anti-infectives, anti-inflammatory drugs, antirheumatics, and antivirals. When users submit a molecule to the server, it suggests possible off-targets, as well as the molecule's similarity to other drugs in the server's library based on interaction profiles.
The researchers believe that the server can complement gene expression profiles that are used to find new indications for drugs.
"Though high-throughput technology such as microarray[s] [have] the potential to generate large quantities of data for analyzing drug-drug associations, this methodology, although robust, can also be costly," they wrote, adding that "reliability and quality measures still need to be improved."
Furthermore, additional targets for drugs, as well as adverse reactions, are sometimes discovered by chance, Lun Yang, a researcher at Shanghai Jiao Tong University and one of the authors of the study, told BioInform, adding that the in silico method provides a more systematic approach to the issue.
When users submit a molecule, the server "calculate[s] the binding energy between the uploaded molecule and the targets," the investigators wrote, and provides "the interaction profile of the [query] drug towards all targets in the database."
Specifically, DRAR-CPI provides "positive or negative association scores between the user’s molecule and our library drugs based on their interaction profiles ... [with the] proteins and will also suggest candidate off-targets that tend to interact with it."
Furthermore, because the library includes treatment indications for each drug, as well as possible adverse reactions, "users can predict potential indications or ADRs based on the association scores of their molecule."
Like a search engine, DRAR-CPI "will suggest the most relevant drug that has the most similar fingerprint with [the] query drug," Yang explained. "If we know the adverse drug reaction or the drug indication of the library drug, we could predict whether the [query drug] has the same kind of therapeutic effect like the drug [in the] library."
DRAR-CPI compares query drugs to a library of 385 proteins derived from 353 proteins in UniProt and 254 small molecules culled from 166 molecules in DrugBank with known descriptions.
The team uses the Dock software to score interactions between the targets and the library drugs and to score interactions between query drugs and protein targets.
The docking scores are transformed into a matrix of "Z-scores," where targets with values greater than one are treated as unfavorable targets while those with values less than minus one are considered favorable.
These numbers are then used to compute an "enrichment score" for both favorable and unfavorable targets that’s used to calculate the final association score, S.
DRAR-CPI takes between six and 20 hours to run one drug molecule and provides users via e-mail with a list of drugs that have a similar interaction profile with the query molecule as well as information on known indications and adverse reactions and possible off-targets that may interact with the user's molecule.
To be sure their results were accurate, the team compared DRAR-CPI's predictions with data from gene-expression profiles of 12 drugs.
Out of 87 associations made by the server, the investigators reported that 64 associations, or 74 percent, matched the results from the gene expression experiments.
Yang explained that it is likely the remaining 26 percent of the associations are not based on direct drug-protein interactions but on other aspects such as gene or protein expression, which he admitted is a shortcoming of the method.
In one sample case study described in the paper, the team used the server to find intentional indications and adverse events for rosigliazone, an anti-diabetic that activates peroxisome proliferator-activated receptors, or PRARs, to reduce resistant to insulin. The drug is marketed by GlaxoSmithKline as Avandia.
They found that fulvestrant, an anti-estrogenic drug that’s used to treat hormone receptor positive metastatic breast cancer, had the closest similarity to rosigliazone with an association score of one.
Because PRARs are linked to human primary and metastatic breast cancer, the team inferred that rosigliazone could potentially be a new treatment for the cancer.
In a separate paper published in March in PLoS Computational Biology, Yang and colleagues used their approach to identify a gene that increases an individual's susceptibility for clozapine-induced agranulocytosis, or CIA.
Clozapine, marketed by Novartis as Clozaril, treats schizophrenia but is associated with agranulocytosis — a condition that lowers the white blood cell count leaving victims open to infections. A similar drug, olanzapine, marketed as Zyprexa by Eli Lilly, has a lower incidence of the disease.
The team used the chemical-protein docking approach behind DRAR-CPI along with microarray experiments to identify HSPA1A, a known susceptibility gene for CIA, as an off-target for clozapine but not olanzapine.
Moving forward, Yang's team plans to continue building its library of human proteins as well as regulatory approved drugs.
Reusing Is In
In the Nucleic Acids Research paper, the team stated the oft-cited statistic that more than 90 percent of new drug candidates fail during the development process. Furthermore, the latest estimates from the Tufts Center for the Study of Drug Development put the average cost of developing a new drug at around $1.3 billion — a situation that makes drug repositioning a promising option for many drug makers.
Well-known examples of repurposed drugs cited in the paper include sildenafil citrate, marketed as Viagra by Pfizer, and raloxifene, marketed as Evista by Eli Lilly, which were both originally developed to treat diseases other than their current primary indications.
Other efforts in the academic space to support drug repurposing include a system called Promiscuous developed by researchers at Charité University Medicine Berlin and the University of California, San Diego, which contains 25,000 drugs, including withdrawn and experimental therapies, annotated with drug-protein and protein-protein relationships.
UCSD's Phil Bourne, a developer of Promiscuous, has also co-developed Off-Target Pipeline, a computational method that combines molecular dynamics simulation, free energy calculations, ligand binding site comparison, and biological network analysis to identify putative off-target effects of lead compounds.
Commercially, companies such as Entelos are attempting to use computational approaches to find new uses for existing drugs. In 2007, the company said, among other goals, that it planned to use its PhysioLab biosimulation platform for drug repositioning (BI 04/20/2007).
More recently, at the annual meeting of the American Association for Cancer Research in Orlando, Fla., Life Biosystems said that its Molecular Analysis of Side Effect information, or MASE, system, which is used to predict the outcomes of specific combinations of drugs, targets, and metabolizing enzymes, shows promise for drug repositioning (BI 04/15/2011).
Have topics you'd like to see covered in BioInform? Contact the editor at uthomas [at] genomeweb [.] com.